Nobody wants to talk about taxonomy. It's not exciting. It doesn't make good slides. No brand manager has ever said "I'm thrilled about the SKU mapping."
But taxonomy is the foundation that every retail media claim stands on. If products and categories don't map cleanly, targeting lies and reporting lies. Trust disappears fast, not because someone cheated, but because the plumbing was wrong.
What taxonomy means in retail media
Taxonomy is the classification system that organizes products into categories, subcategories, and segments. It's how the retailer's product catalog is structured, and by extension, how every audience, every targeting rule, and every campaign report is organized.
A simple example: a protein bar. Is it classified as: - Confectionery → Bars → Protein bars? - Health & Wellness → Sports Nutrition → Protein bars? - Snacks → Cereal bars → Protein bars?
The answer depends on the retailer's taxonomy. And the answer matters, because:
If a brand targets "Health & Wellness shoppers," the protein bar buyer is included, or not, based on where the product sits in the taxonomy.
If a campaign report measures "uplift in the Snacks category," the protein bar's sales are included, or not, based on classification.
If a category affinity model identifies "shoppers with Health & Wellness affinity," the signal quality depends on whether health products are correctly classified.
One misclassification doesn't break anything. A thousand misclassifications break everything.
SKU mapping: the granular layer
Below taxonomy sits SKU mapping, the linkage between individual product codes and the taxonomy tree.
A single product might exist under different SKUs across different stores, different pack sizes, or different promotional variants. A 200g yogurt and a 500g yogurt might be different SKUs but the same product line. A promotional bundle might have its own SKU that doesn't map neatly to any single category.
SKU mapping is the work of connecting every individual product code to the right place in the taxonomy. It's unglamorous, tedious, and absolutely critical.
When we tell retailers we need ePOS store-level and SKU-level data, this is part of what we mean. Not just the transactions, but transactions with clean, consistent product identification that can be mapped to a coherent category structure.
Where taxonomy breaks targeting
Targeting breaks when the taxonomy doesn't match the brand's view of their competitive set.
A beverage brand might define their competitive set as "functional beverages", energy drinks, sports drinks, enhanced water, kombucha. But the retailer's taxonomy might split these across: - Beverages → Soft Drinks → Energy Drinks - Beverages → Water → Enhanced Water - Beverages → Chilled → Kombucha - Health → Sports Nutrition → Sports Drinks A campaign targeting "functional beverage buyers" needs to pull from four different taxonomy branches. If the platform only allows targeting by single category, the audience is incomplete. If the platform allows custom groupings but the SKU mapping is inconsistent across stores, the audience is noisy.
This is a solvable problem, but it requires intentional taxonomy management, not just defaulting to whatever the retailer's ERP system produces.
Where taxonomy breaks measurement
Measurement breaks even more visibly.
A campaign report says: "8% uplift in the Dairy category." But the Dairy taxonomy includes milk, cheese, yogurt, cream, butter, and dairy alternatives. The campaign promoted yogurt. Did the whole category lift, or just yogurt? If just yogurt, is the 8% an accurate reflection, or is it diluted by the broader category denominator?
The opposite problem: the taxonomy is too granular. The report measures "uplift in Greek yogurt, 150g single-serve, fruit flavor." The sample size for that specific SKU subset is too small to reach statistical significance. The report shows "no significant uplift" even though the campaign worked, it just can't be proven at that granularity level.
Getting the measurement taxonomy right means matching the level of aggregation to the campaign's scope and the statistical requirements of the methodology. That's a design decision, not a default.
The retailer's taxonomy vs the brand's taxonomy
Here's a persistent friction point: the retailer organizes products by how they stock shelves. The brand organizes products by how they go to market.
The retailer's taxonomy is designed for store operations, planograms, shelf sets, ordering systems. The brand's taxonomy is designed for marketing, brand families, product lines, competitive sets, consumer need states.
These don't always align. And retail media sits in the middle, needing to speak both languages.
The platform's job is to provide a translation layer, allowing brands to define audiences and measurement in their own terms while mapping to the retailer's underlying data structure. Custom audience definitions, flexible reporting hierarchies, and the ability to group SKUs across taxonomy branches are all necessary features.
The governance imperative
Taxonomy isn't a one-time setup. Products launch, get delisted, get reformulated, change pack sizes, move categories. Without ongoing governance, the taxonomy degrades over time.
A new product launches without a category assignment. It sits in "Uncategorized" for three months. During those three months, every campaign that targets its category misses it, every report that measures its category excludes it, and every affinity model that should include it doesn't.
Taxonomy governance means: new products are classified at launch. Reclassifications are tracked and versioned. Cross-retailer mappings are maintained (because the same product has different codes at different retailers). Anomalies are flagged and fixed.
This is operational work, not data science, not AI, just disciplined product data management. And it's non-negotiable for any RMN that wants its numbers to be trusted.
The bottom line
Taxonomy and SKU mapping are the hidden plumbing of retail media. If they're clean, everything built on top, targeting, measurement, audiences, reporting, works as promised. If they're messy, the numbers lie, the targeting misfires, and trust erodes.
No one gets excited about taxonomy. But every credible retail media result depends on it. It's the boring infrastructure that makes the exciting promises possible.
If your retail media reports are questioned, check the taxonomy before you check the model. The plumbing is usually where the leak starts.
Related Reading
- Predictive Audiences: When Retail Media Moves From Segments to Probability
- Verified Impressions: The Difference Between Served and Proven
- Retail Signals: The Raw Material That Makes Retail Media Perform
- Category Affinity: The Shortcut to Relevant Targeting
- True Reach: What Reach Actually Means
Ready to see how this works in practice?
Footprints AI helps brands and retailers measure what matters. See our customer success stories or get in touch to discuss your retail media strategy.
